arXiv subjects

Yuan Ji

Publications and source records attributed to Yuan Ji.

At least 55 records · Page 3Linked to original sources

arXiv2019

Reversible Data Hiding in JPEG Images with Multi-objective Optimization

Among various methods of reversible data hiding (RDH) in JPEG images, the consideration in designing is only the image quality, but the image quality and the file size expansion are equally important in JPEG images. Based on this situation, we propose a RDH scheme in JPEG images considering both the image quality and the file size expansion while designing the algorithm. The multi-objective optimization strategy is utilized to realize the balance of the two objectives. Specifically, the cover is divided into several non-overlapping signals firstly, and after that, the embedding costs of signals are calculated using the knowledge of the JPEG compression. Next, the optimized combination of signals for embedding data is gained by the multi-objective optimization. Experimental results show the better performance of our proposed RDH compared with state-of-the-art RDH in JPEG images.

cs.MM↗

arXiv2019

The i3+3 Design for Phase I Clinical Trials

Purpose: The 3+3 design has been shown to be less likely to achieve the objectives of phase I dose-finding trials when compared with more advanced model-based designs. One major criticism of the 3+3 design is that it is based on simple rules, does not depend on statistical models for inference, and leads to unsafe and unreliable operating characteristics. On the other hand, being rule-based allows 3+3 to be easily understood and implemented in practice, making it the first choice among clinicians. Is it possible to have a rule-based design with great performance? Methods: We propose a new rule-based design called i3+3, where the letter "i" represents the word "interval". The i3+3 design is based on simple but more advanced rules that account for the variabilities in the observed data. We compare the operating characteristics for the proposed i3+3 design with other popular phase I designs by simulation. Results: The i3+3 design is far superior than the 3+3 design in trial safety and the ability to identify the true MTD. Compared with model-based phase I designs, i3+3 also demonstrates comparable performances. In other words, the i3+3 design possesses both the simplicity and transparency of the rule-based approaches, and the superior operating characteristics seen in model-based approaches. An online R Shiny tool (https://i3design.shinyapps.io/i3plus3/) is provided to illustrate the i3+3 design, although in practice it requires no software to design or conduct a dose-finding trial. Conclusion: The i3+3 design could be a practice-altering method for the clinical community.

stat.ME↗

arXiv2019

Bayesian Double Feature Allocation for Phenotyping with Electronic Health Records

We propose a categorical matrix factorization method to infer latent diseases from electronic health records (EHR) data in an unsupervised manner. A latent disease is defined as an unknown biological aberration that causes a set of common symptoms for a group of patients. The proposed approach is based on a novel double feature allocation model which simultaneously allocates features to the rows and the columns of a categorical matrix. Using a Bayesian approach, available prior information on known diseases greatly improves identifiability and interpretability of latent diseases. This includes known diagnoses for patients and known association of diseases with symptoms. We validate the proposed approach by simulation studies including mis-specified models and comparison with sparse latent factor models. In the application to Chinese EHR data, we find interesting results, some of which agree with related clinical and medical knowledge.

stat.AP↗

arXiv2018

Scalable Bayesian Nonparametric Clustering and Classification

We develop a scalable multi-step Monte Carlo algorithm for inference under a large class of nonparametric Bayesian models for clustering and classification. Each step is "embarrassingly parallel" and can be implemented using the same Markov chain Monte Carlo sampler. The simplicity and generality of our approach makes inference for a wide range of Bayesian nonparametric mixture models applicable to large datasets. Specifically, we apply the approach to inference under a product partition model with regression on covariates. We show results for inference with two motivating data sets: a large set of electronic health records (EHR) and a bank telemarketing dataset. We find interesting clusters and favorable classification performance relative to other widely used competing classifiers.

stat.CO↗

arXiv2018

Heterogeneous Reciprocal Graphical Models

We develop novel hierarchical reciprocal graphical models to infer gene networks from heterogeneous data. In the case of data that can be naturally divided into known groups, we propose to connect graphs by introducing a hierarchical prior across group-specific graphs, including a correlation on edge strengths across graphs. Thresholding priors are applied to induce sparsity of the estimated networks. In the case of unknown groups, we cluster subjects into subpopulations and jointly estimate cluster-specific gene networks, again using similar hierarchical priors across clusters. We illustrate the proposed approach by simulation studies and two applications with multiplatform genomic data for multiple cancers.

stat.ME↗

arXiv2017

TreeClone: Reconstruction of Tumor Subclone Phylogeny Based on Mutation Pairs using Next Generation Sequencing Data

We present TreeClone, a latent feature allocation model to reconstruct tumor subclones subject to phylogenetic evolution that mimics tumor evolution. Similar to most current methods, we consider data from next-generation sequencing of tumor DNA. Unlike most methods that use information in short reads mapped to single nucleotide variants (SNVs), we consider subclone phylogeny reconstruction using pairs of two proximal SNVs that can be mapped by the same short reads. As part of the Bayesian inference model, we construct a phylogenetic tree prior. The use of the tree structure in the prior greatly strengthens inference. Only subclones that can be explained by a phylogenetic tree are assigned non-negligible probabilities. The proposed Bayesian framework implies posterior distributions on the number of subclones, their genotypes, cellular proportions, and the phylogenetic tree spanned by the inferred subclones. The proposed method is validated against different sets of simulated and real-world data using single and multiple tumor samples. An open source software package is available at http://www.compgenome.org/treeclone.

stat.AP↗

arXiv2017

A Bayesian Mixture Model for Clustering on the Stiefel Manifold

Analysis of a Bayesian mixture model for the Matrix Langevin distribution on the Stiefel manifold is presented. The model exploits a particular parametrization of the Matrix Langevin distribution, various aspects of which are elaborated on. A general, and novel, family of conjugate priors, and an efficient Markov chain Monte Carlo (MCMC) sampling scheme for the corresponding posteriors is then developed for the mixture model. Theoretical properties of the prior and posterior distributions, including posterior consistency, are explored in detail. Extensive simulation experiments are presented to validate the efficacy of the framework. Real-world examples, including a large scale neuroimaging dataset, are analyzed to demonstrate the computational tractability of the approach.

stat.ME↗

arXiv2017

On the Interval-Based Dose-Finding Designs

The landscape of dose-finding designs for phase I clinical trials is rapidly shifting in the recent years, noticeably marked by the emergence of interval-based designs. We categorize them as the iDesigns and the IB-Designs. The iDesigns are originated by the toxicity probability inter- val (TPI) designs and its two modifications, the mTPI and mTPI-2 designs. The IB-Designs started as the cumulative cohort design (CCD) and is recently extended by the BOIN design. We discuss the differences and similarities between these two classes of interval-based designs, and compare their simulation performance with popular non-interval designs, such as the CRM and 3+3 designs. We also show that in addition to the population-level operating characteristics from simulated trials, investigators should also assess the dose-finding decision tables from the implemented designs to better understand the per-trial and per-patient behavior. This is particularly important for nonstatisticians to assess the designs with transparency. We pro- vide, to our knowledge, the most comprehensive simulation-based comparative study on various interval-based dose-finding designs.

stat.ME↗

arXiv2017

AAA: Triple-adaptive Bayesian designs for the identification of optimal dose combinations in dual-agent dose-finding trials

We propose a flexible design for the identification of optimal dose combinations in dual-agent dose-finding clinical trials. The design is called AAA, standing for three adaptations: adaptive model selection, adaptive dose insertion, and adaptive cohort divi- sion. The adaptations highlight the need and opportunity for innovation for dual-agent dose finding, and are supported by the numerical results presented in the proposed simulation studies. To our knowledge, this is the first design that allows for all three adaptations at the same time. We find that AAA improves the statistical inference, enhances the chance of finding the optimal dose combinations, and shortens the trial duration. A clinical trial is being planned to apply the AAA design.

stat.ME↗

arXiv2017

Optimal control-based inverse determination of electrode distribution for electroosmotic micromixer

This paper presents an optimal control-based inverse method used to determine the distribution of the electrodes for the electroosmotic micromixers with external driven flow from the inlet. Based on the optimal control method, one Dirichlet boundary control problem is constructed to inversely find the optimal distribution of the electrodes on the sidewalls of electroosmotic micromixers and achieve the acceptable mixing performance. After solving the boundary control problem, the step-shaped distribution of the external electric potential imposed on the sidewalls can be obtained and the distribution of electrodes can be inversely determined according to the obtained external electric potential. Numerical results are also provided to demonstrate the effectivity of the proposed method.

physics.flu-dyn↗

arXiv2017

PairClone: A Bayesian Subclone Caller Based on Mutation Pairs

Tumor cell populations can be thought of as being composed of homogeneous cell subpopulations, with each subpopulation being characterized by overlapping sets of single nucleotide variants (SNVs). Such subpopulations are known as subclones and are an important target for precision medicine. Reconstructing such subclones from next-generation sequencing (NGS) data is one of the major challenges in precision medicine. We present PairClone as a new tool to implement this reconstruction. The main idea of PairClone is to model short reads mapped to pairs of proximal SNVs. In contrast, most existing methods use only marginal reads for unpaired SNVs. Using Bayesian nonparametric models, we estimate posterior probabilities of the number, genotypes and population frequencies of subclones in one or more tumor sample. We use the categorical Indian buffet process (cIBP) as a prior probability model for subclones that are represented as vectors of categorical matrices that record the corresponding sets of mutation pairs. Performance of PairClone is assessed using simulated and real datasets. An open source software package can be obtained at http://www.compgenome.org/pairclone.

stat.AP↗

arXiv2016

A Bayesian Interval Dose-Finding Design Addressing Ockham's Razor: mTPI-2

There has been an increasing interest in using interval-based Bayesian designs for dose finding, one of which is the modified toxicity probability interval (mTPI) method. We show that the decision rules in mTPI correspond to an optimal rule under a formal Bayesian decision theoretic framework. However, the probability models in mTPI are overly sharpened by the Ockham's razor, which, while in general helps with parsimonious statistical inference, leads to suboptimal decisions in small-sample inference such as dose finding. We propose a new framework that blunts the Ockham's razor, and demonstrate the superior performance of the new method, called mTPI-2. An online web tool is provided for users who can generate the design, conduct clinical trials, and examine operating characteristics of the designs through big data and crowd sourcing.

stat.ME↗

arXiv2016

Reciprocal Graphical Models for Integrative Gene Regulatory Network Analysis

Constructing gene regulatory networks is a fundamental task in systems biology. We introduce a Gaussian reciprocal graphical model for inference about gene regulatory relationships by integrating mRNA gene expression and DNA level information including copy number and methylation. Data integration allows for inference on the directionality of certain regulatory relationships, which would be otherwise indistinguishable due to Markov equivalence. Efficient inference is developed based on simultaneous equation models. Bayesian model selection techniques are adopted to estimate the graph structure. We illustrate our approach by simulations and two applications in ZODIAC pairwise gene interaction analysis and colon adenocarcinoma pathway analysis.

stat.ME↗

arXiv2016

An Ensemble EM Algorithm for Bayesian Variable Selection

We study the Bayesian approach to variable selection in the context of linear regression. Motivated by a recent work by Rockova and George (2014), we propose an EM algorithm that returns the MAP estimate of the set of relevant variables. Due to its particular updating scheme, our algorithm can be implemented efficiently without inverting a large matrix in each iteration and therefore can scale up with big data. We also show that the MAP estimate returned by our EM algorithm achieves variable selection consistency even when $p$ diverges with $n$. In practice, our algorithm could get stuck with local modes, a common problem with EM algorithms. To address this issue, we propose an ensemble EM algorithm, in which we repeatedly apply the EM algorithm on a subset of the samples with a subset of the covariates, and then aggregate the variable selection results across those bootstrap replicates. Empirical studies have demonstrated the superior performance of the ensemble EM algorithm.

stat.CO↗

arXiv2015

A Bayesian feature allocation model for tumor heterogeneity

We develop a feature allocation model for inference on genetic tumor variation using next-generation sequencing data. Specifically, we record single nucleotide variants (SNVs) based on short reads mapped to human reference genome and characterize tumor heterogeneity by latent haplotypes defined as a scaffold of SNVs on the same homologous genome. For multiple samples from a single tumor, assuming that each sample is composed of some sample-specific proportions of these haplotypes, we then fit the observed variant allele fractions of SNVs for each sample and estimate the proportions of haplotypes. Varying proportions of haplotypes across samples is evidence of tumor heterogeneity since it implies varying composition of cell subpopulations. Taking a Bayesian perspective, we proceed with a prior probability model for all relevant unknown quantities, including, in particular, a prior probability model on the binary indicators that characterize the latent haplotypes. Such prior models are known as feature allocation models. Specifically, we define a simplified version of the Indian buffet process, one of the most traditional feature allocation models. The proposed model allows overlapping clustering of SNVs in defining latent haplotypes, which reflects the evolutionary process of subclonal expansion in tumor samples.

stat.AP↗

arXiv2014

MAD Bayes for Tumor Heterogeneity Feature Allocation with Non-Normal Sampling

We propose small-variance asymptotic approximations for the inference of tumor heterogeneity (TH) using next-generation sequencing data. Understanding TH is an important and open research problem in biology. The lack of appropriate statistical inference is a critical gap in existing methods that the proposed approach aims to fill. We build on a hierarchical model with an exponential family likelihood and a feature allocation prior. The proposed approach generalizes similar small-variance approximations proposed by Kulis and Jordan (2012) and Broderick et.al (2012) for inference with Dirichlet process mixture and Indian buffet prior models under normal sampling. We show that the new algorithm can successfully recover latent structures of different subclones and is also magnitude faster than available Markov chain Monte Carlo samplers, the latter often practically infeasible for high-dimensional genomics data. The proposed approach is scalable, simple to implement and benefits from the flexibility of Bayesian nonparametric models. More importantly, it provides a useful tool for the biological community for estimating cell subtypes in tumor samples.

stat.ME↗

arXiv2014

Bayesian sparse graphical models for classification with application to protein expression data

Reverse-phase protein array (RPPA) analysis is a powerful, relatively new platform that allows for high-throughput, quantitative analysis of protein networks. One of the challenges that currently limit the potential of this technology is the lack of methods that allow for accurate data modeling and identification of related networks and samples. Such models may improve the accuracy of biological sample classification based on patterns of protein network activation and provide insight into the distinct biological relationships underlying different types of cancer. Motivated by RPPA data, we propose a Bayesian sparse graphical modeling approach that uses selection priors on the conditional relationships in the presence of class information. The novelty of our Bayesian model lies in the ability to draw information from the network data as well as from the associated categorical outcome in a unified hierarchical model for classification. In addition, our method allows for intuitive integration of a priori network information directly in the model and allows for posterior inference on the network topologies both within and between classes. Applying our methodology to an RPPA data set generated from panels of human breast cancer and ovarian cancer cell lines, we demonstrate that the model is able to distinguish the different cancer cell types more accurately than several existing models and to identify differential regulation of components of a critical signaling network (the PI3K-AKT pathway) between these two types of cancer. This approach represents a powerful new tool that can be used to improve our understanding of protein networks in cancer.

stat.ME↗

arXiv2014

Bayesian Inference for Tumor Subclones Accounting for Sequencing and Structural Variants

Tumor samples are heterogeneous. They consist of different subclones that are characterized by differences in DNA nucleotide sequences and copy numbers on multiple loci. Heterogeneity can be measured through the identification of the subclonal copy number and sequence at a selected set of loci. Understanding that the accurate identification of variant allele fractions greatly depends on a precise determination of copy numbers, we develop a Bayesian feature allocation model for jointly calling subclonal copy numbers and the corresponding allele sequences for the same loci. The proposed method utilizes three random matrices, L, Z and w to represent subclonal copy numbers (L), numbers of subclonal variant alleles (Z) and cellular fractions of subclones in samples (w), respectively. The unknown number of subclones implies a random number of columns for these matrices. We use next-generation sequencing data to estimate the subclonal structures through inference on these three matrices. Using simulation studies and a real data analysis, we demonstrate how posterior inference on the subclonal structure is enhanced with the joint modeling of both structure and sequencing variants on subclonal genomes. Software is available at http://compgenome.org/BayClone2.

stat.ME↗